Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Im confused about what to do. I was thinking of running two seperate multiple regressions with a DV in each.. its after this, that I'm stuck. How do I see what effect the two DV's have combined? Or am I going about the whole thing wrong. Help!

My IV's are gender and group.

My DV's will be scores on two seperate psychometric tests (likert scales)

I hope to have at least 100 people in each group (3 groups) so sample size will be roughly 300.

share|improve this question
+1 for thinking about this before collecting data. Good on you! I already answered that MANOVA ( might be the way to go, but this presupposes normally distributed residuals, which your Likert scales will not provide, so I deleted my non-answer right away... – Stephan Kolassa Nov 29 '12 at 18:01
Thanks anyway Stephan, any help is appreciated!! :) – karen murray Nov 29 '12 at 18:06
Googling for "alternative to MANOVA for ordinal data" did not turn up a lot of helpful things, except for perhaps this: You may want to look into that paper, perhaps there are a few pointers there. – Stephan Kolassa Nov 29 '12 at 18:07
I'll have a look now, thanks a mill! – karen murray Nov 29 '12 at 18:12
Perhaps if you think through what you mean by "the effect the two DVs have combined" it will help decide if you need to worry about this or you can just run to separate regressions. Generally you aren't thinking about the "effect" of a DV. So in what way is it important to look at the two of them simultaneously? This will depend on your research question. – Peter Ellis Nov 29 '12 at 18:47

Are you interested in examining the correlation between the dependent variables in the same model? I can't speak to the multiple independent variables part of the question, but you could investigate using a linear mixed model with multiple response variables (if your data will be longitudinal). I don't know of a website (I'm sure there is stuff out there, I just don't have a reference), but the book by Jeffrey Long, 'Longitudinal data analysis for the behavioral sciences using R' may be of use. Chapter 13 (p 501) has a section on models with multiple dv's.

share|improve this answer

You probably know about this, but I often start at this point:

share|improve this answer
Hi RLang, Thats a really helpful page, thank you. Although my data is ordinal and there is no option for that there. Maybe I can use the non parametric equivalent of a multivariate multiple linear regression, whatever that is.. – karen murray Nov 29 '12 at 22:09

You said "How do I see what effect the two DV's have combined". So, I believe your two DV's are somehow related based on theory. In this case, running independent equations each for one DV is totally wrong! Imagine you want to measure "speaking skill" and one of your DV's is "accuracy of talking" and the other one is "fluency of talking". In this respect, theory says that fluency and accuracy are related and are two sub-dimensions of the same concept-speaking skill. In this case, the two DV's should be considered together. Whenever it is the case, you have a continuous DV and categorical IV(s), Multivariate ANOVA (MANOVA) should be used. MANOVA is a variant of ANOVA that can incorporate multiple continuous DV's. Also, the number of IV's is irrelevant here as both ANOVA and MANOVA can accommodate 1 IV or 2 IV's or 3 IV's, etc. (called one way, two way, three-way etc. ANOVA or MANOVA)

Also, pay attention that MANOVA is a parametric test, meaning that your measurement must be continuous. it is hard to assume that measurement by any Likert-type scale with less than 7 anchors is continuous.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.